A Scalable Cybersecurity Defense Network for Adaptive Intrusion Detection Using Machine Learning Techniques

Intrusion Detection Systems (IDS) are essential for network security, but traditional models struggle with adaptability, false positives, and inefficient feature selection. This research proposes a Scalable Cybersecurity Defense Network (SCDN) that integrates dynamic feature selection techniques and advanced machine learning models to enhance classification accuracy and adaptability. The Selective Attribute Reduction Technique (SART) optimizes feature selection by combining filter-based, wrapper-based, and embedded methods, ensuring dimensionality reduction without compromising performance. Additionally, the Adaptive Stacked Classifier Network (ASCN) leverages a hybrid ensemble learning approach, utilizing Random Forest (RF), Support Vector Machines (SVM), and Logistic Regression (LR) as base classifiers, with AdaBoost as the meta-classifier. The proposed method is initially designed using the online CIC IoT Dataset 2023, achieving an accuracy of 99.55%, precision of 99.55%, recall of 99.55%, F1-score of 99.55%, and an error rate of 0.45%. To validate real-world applicability, the model is further tested on a simulation-generated dataset, obtained through the NS2 simulation process. The simulation results indicate an accuracy of 99.28%, precision of 99.29%, recall of 99.28%, F1-score of 99.28%, and an error rate of 0.72%, demonstrating the model’s effectiveness in real-time network environments. The proposed SCDN framework significantly enhances intrusion detection performance by reducing false positives and adapting to emerging cybersecurity threats.